What Does #Freedom Look Like? Instagram and the Visual Imagination of Values

Instagram is the place for the visualization of everything, from travel and food to abstract concepts such as freedom. Over the past decade, the platform has introduced a bottom-up process where users co-produce image repertoires that shape the boundaries of the imaginable. Drawing on an epistemology of social constructionism, we ask which visual repertoires are associated with value-related terms on Instagram. We studied 20 widely used value hashtags, sampling the top 100 posts for each (N ¼ 2,000). A combined qualitative–quantitative content analysis revealed that 19 of the 20 hashtags possess distinct visual footprints, typically reflecting an orientation toward the self and an emphasis on consumption. We conclude by discussing three implications of our findings: the role of images in the social construction of the meaning of values, the distinction between internalized and externalized value depictions, and aestheticized consumption as an organizing principle of Instagram’s mainstream.


Introduction
Over the past decade, Instagram has pushed the visual to the center of our social media experience. From the perfect avocado toast to post-Soviet Brutalist architecture, the platform serves as a place for the visualization of everything (Leaver, Highfield, & Abidin, 2020). This "everything" also includes abstract concepts such as faith, happiness, or freedom. To be sure, visualizing the abstract is not a new phenomenon; traditional iconography (Panofsky, 1955) and stock photography (Frosh, 2003) offer poignant historical examples. However, Instagram transforms the process of visualizing the abstract into a bottom-up, collaborative endeavor by organizing images around user-generated searchable hashtags. Theoretically, every one of Instagram's one billion users can contribute to this process, generating seemingly unlimited possibilities. In practice, however, visualization is inevitably constrained by both technological features and socio-cultural processes (Highfield & Leaver, 2016;Manovich, 2016). As wide as it might be, any visual repertoire on Instagram is ultimately bounded.
In this paper, we adopt a social constructionist epistemology to investigate Instagram as a photographic archive (Sekula, 1989), understood as a set of logics that classify images into categories and encourage the reproduction of existing visual patterns. One of the dominant archival logics on Instagram is social tagging. Hashtags organize the generation and circulation of new images on the platform. They also organize users in so-called "ambient" communities that constantly negotiate the meaning of their shared values (Zappavigna & Martin, 2018). Generated by users, hashtag categories function as a bottom-up "folk taxonomy" or "folksonomy" (Highfield & Leaver, 2014) that contributes to the socially constructed meaning of concrete issues and abstract concepts.
We explore the notion of user-generated image archives through a particular set of abstract concepts: values. Integrating existing theories, we define values as guiding beliefs about the desirable that relate to people's perceptions of how they should conduct their lives, both as individuals and as collectives (Inglehart, 1971;Schwartz, 1992). Values are a fundamental part of being human. Perceptions about good and bad both shape and are shaped by everyday life. Yet the basic meaning of values is often disputed and always subject to negotiation. Visual communication only increases the ambiguity of values, since the visualization of abstract concepts is particularly open to interpretation (Hallinan, Scharlach, & Shifman, 2021). Thus, popular images associated with values can define the boundaries of what people imagine as happiness, faith, or equality.
While Instagram has been broadly identified as a place where images and values influence each other (Abidin, 2016;Baker & Walsh, 2018), the concept of values has not been foregrounded in studies of hashtags on the platform. To be sure, studies of politically charged hashtags such as #BlackLivesMatter explore issues that are heavily infused with values (Wilkins, Livingstone, & Levine, 2019). However, these hashtags tend to be both specific (as they refer to political causes or movements) and vague (as their link with values is often left implicit). An explicit focus on the visualization of value hashtags will allow us to look beyond specific contexts and investigate the social construction of values in digital spheres.
Therefore, we set off to investigate: which visual repertoires are associated with value-related terms on Instagram?. Our research question can also be phrased as: "what does [#value] look like?" We first survey three conceptual domains leading to our study: the notion of Instagram as a user-generated archive, the relationship between values and visuality, and value-related issues associated with Instagram. We then describe our methods of analysis, which combine qualitative and quantitative investigations of a corpus of 2,000 Instagram posts associated with 20 value-related hashtags from "mainstream" Instagram (i.e., public posts with high visibility on the platform). In the "Results" section, we survey the unique visual footprint associated with each of the value hashtags. Next, we present a holistic qualitative analysis organized around two sets of axes (Figures 3 and 4), one concerned with the content of images and the other with the interpretation of their socio-cultural significance. Finally, we discuss the implications of our study for the relationship between values and visualization on Instagram.
Instagram as a user-generated archive for the visualization of everything The epistemological orientation guiding this study is that of social constructionism. One of the basic premises of this approach is that reality is not given; instead, people construct reality as they collectively interpret their surrounding environment and engage in communication to describe it (Berger & Luckmann, 1966). Simply put, social constructionism locates the meaning of any phenomenon in what we say about it. Accordingly, we conceptualize value terms as floating signifiers, the meaning of which is unstable and open to interpretation. In the context of Instagram, images operate as sites for the production of meaning through social interactions. Rather than objectively reproducing their referents, visual utterances contribute to the social signification of the referents they claim to represent (Foucault, 1972). As representations accrue over time, they coalesce into visual paradigms that influence future representations of the same referents. In accordance with Giddens' (1986) notions of structuration, our main claim is that people produce images, the accumulation of images produces structures, and these structures constrain what images people can and do produce thereafter. Of course, the process of structuration is circular and has no "time zero"; the images that people produce are always influenced by structures, including the features of the platform itself.
In line with the constructionist epistemology guiding this work, we conceptualize images as both objects and agents of classification (Frosh, 2003, p. 91). Images can be categorized according to a wide range of criteria (e.g., context, content, and style) and they can be used to categorize people and objects (e.g., passports, mug shots, catalogs, and advertisements). This dual notion is inspired by the Foucauldian conceptualization of the archive (Foucault, 1972). For Foucault, the archive is a set of rules that groups together content with a certain degree of regularity, promotes the production of new entries that resemble previously archived ones, and is flexible enough to allow for transformation over time. Applying this notion to images, it is possible to speak of the archive as both an actual repository where pictures are stored and as a stylistic paradigm that favors standardization (Sekula, 1989). In short, images exist as objects stored in actual archives and are produced according to rules favoring regularity.
This dual perception of the archive is particularly useful for interrogating Instagram, which undeniably shares many of the material qualities of the archive. It is a place where images are stored and categorized according to several logics. Some of these categorizations come directly from authors (e.g., hashtags), some emerge through interaction between users on the platform (e.g., comments and likes), and some are inscribed in the platform through its Terms of Service, Community Guidelines, and commercial character (e.g., sponsored content) (Gillespie, 2018). Instagram is, therefore, an archive that favors established organizational principles such as authorship and sponsorship, but also new ones such as interactivity and searchability.
To a large extent, Instagram's taxonomic archive is organized around hashtags. Produced by users rather than platforms, hashtags are a bottom-up alternative to the top-down organization typical of traditional taxonomies (Noruzi, 2006). Hashtagging is a form of folksonomy that dates back to the early days of the Web and was popularizedat the start of the 2000s on platforms like Flickr and Del.icio.us. The practicebegan on Twitter as a social convention to signal that a post was part of a broader topic (Bruns & Burgess, 2015;Burgess & Baym, 2020) and was eventually formalized in the architecture of most commercial social media platforms (Halavais, 2013). While the use of hashtags can be platform-specific (Highfield & Leaver, 2014), hashtagging is a key affordance that infuses social media with the logic of the archive by making content searchable (Boyd, 2011).
The visual paradigms emerging from this general trend have been described through a number of key concepts in the field. Gibbs, Meese, Arnold, Nansen, and Carter (2015) outline an Instagram-specific platform vernacular composed of the styles and genres that are influenced by the affordances of sharing, filtering, and tagging (p. 258). Manovich (2016) proposes "Instagramism" as a general term to indicate an aesthetic that rejects the need to tell a story or show conventional subjects in favor of images that are "atmospheric, visually perfect, emotional without being aggressive, and subtle as opposed to dramatic" (p. 81). More recently, Leaver et al. (2020) suggest that Instagram favors a set of normalized aesthetics, most of which can be traced back to some influential moments in the platform's history. At its launch in 2010, the affordances of Instagram were inspired by a nostalgia for the Polaroid, as reflected in the logo design and the square format for images. The platform also favored a retro aesthetic that was immediately accessible through filters. Instagram's acquisition by Facebook in 2012 radically changed this ethos, widening the range of affordances available to users. A salient example is the introduction of Stories in 2016, through which the platform definitively moved past its commitment to the 1 Â 1 ratio of the polaroid in favor of a 16 Â 9 ratio suitable for smartphone full-screen visualization.
Instagram's popularity is, therefore, linked to a general transition toward the smartphone as the privileged means of amateur photography (Gómez Cruz & Meyer, 2012). As people started carrying smartphone cameras with them at all times, photography became more mundane. In turn, digital affordances such as filters became central to a "poetics of everyday life" that augments the salience of banal subjects for publication on social media (Ibrahim, 2015).
While the same rules do not apply to all content, it is safe to say that Instagram favors a wide yet bounded number of tropes, templates, clichés, and aesthetics grounded in the materiality of Instagram photography and the vernacular practices of its users (Leaver et al., 2020). Therefore, we argue that Instagram can be seen as an archive that favors an aesthetic that invites repetition and shapes what is visualizable therein. We explore this general notion with a focus on a specific set of Instagram hashtags: value-related terms.

The visualization of values
As stated in the "Introduction" section, we integrate foundational theories (Inglehart, 1971;Hofstede, 1980;Schwartz, 1992) to define values as guiding beliefs about desirable conduct at both the individual and collective levels. Values are personal attributes that are heavily shaped by socio-cultural forces. Residing between thought and expression (Fischer, 2017;Shifman, 2019), the very existence of values is contingent on their dissemination through actions, texts, and images. The perception and communication of values are deeply entwined-when people produce images that reflect the values in which they believe, they instantiate particular understandings of these values, which further impact other people's perceptions.
Historically, the relationship between abstract concepts such as values and their visual representation has been scrutinized by scholars of iconography, broadly defined as the study and interpretation of images. Panofsky's (1955) foundational work outlines three inter-related levels of analysis: pre-iconographic description of images taken at face value, iconographic analysis of images based on familiarity with their socio-historical context, and iconological interpretation of the "essential tendencies of the human mind" embedded in the representation (pp. 40-41). Thus, we contend that a holistic understanding of the visual representation of values should include both a description of value-laden images and an interpretation of their broader meaning.
While inspired by iconography, our study departs from it in two main ways. Traditionally, iconographic investigation starts with the analysis of motifs and works up toward more abstract levels of interpretation. This progression from motifs to values used to be a pragmatic necessity: researchers could not keyword search for images associated with one or more values. The commercial production of images and their organization in catalogs by advertisement agencies partially paved the way for the investigation of images associated with value-related terms (see Frosh's [2003] work on stock photography). Social media, however, has accelerated this process, making large corpora of images searchable through hashtags. Therefore, the affordances of social media platforms allow us to turn iconography on its head and ask: what motifs are associated with specific value terminology?
Relatedly, our second point of departure from iconography is tied to the role of social media platforms in displacing historical patterns in the production and distribution of cultural works (Jenkins, 2006). For many years, religious, state, and corporate elites held a firm grip on the production of images, de facto deciding what could be regarded as iconic representations of people (Sekula, 1989), historical events (Hariman & Lucaites, 2007), or lifestyles (Frosh, 2003). Despite the absence of a systematic review, it is safe to say that some motifs have repeatedly surfaced and emerged as the dominant representation of particular values. For example, protest art by various national liberation movements regularly adapts "liberty" motifs from Eugène Delacroix's Liberty Leading the People, which depicts a personification of Liberty leading the French people toward the 1830 Revolution (Time, 2019). Similarly, commentators in the West saw echoes of Michelangelo's Pietà, a represention of Mary of Nazareth holding Jesus' body after the crucifixion, in one of the defining shots of the so-called "European migration crisis" featuring a Turkish border guard carrying the body of drowned refugee child Alan Kurdi (Fisk, 2016).
Regardless of popularity, top-down representations of any referent (including values) have always been resisted through bottom-up counter-visualization practices (Mirzoeff, 2011). Thus, the radical impact of social media should not be overstated. Social media users did not invent counter-visuality, just as iconic templates predating social media remain very powerful. For example, Caspar David Friedrich's Wanderer above the Sea of Fog, which depicts a young man looking down on a misty landscape from a mountaintop, is such an iconic image of solitary contemplation that it has become a popular Internet meme (Know Your Meme, 2020). Interestingly, we found this motif as a recurrent feature in our Freedom dataset (as detailed below). However, social media enables a larger-than-ever number of amateur photographers to autonomously circulate their images. In turn, this introduces opportunities for the bottom-up construction of new iconic templates on a scale that is potentially global.
Thus, our study borrows from the toolkit of traditional iconography but looks at the relatively new phenomenon of extremely large-scale bottom-up iconic representation of values, a process that takes place at the interface of platforms, users, and the images they produce. We see the relationship between a value and its associated repertoire of images as mutually reinforcing. Value-related terms "anchor" (Barthes, 1981) the visual representation by suggesting the features it is meant to embed. In turn, the representation offers a visual confirmation that "this is what a value looks like." Accordingly, value-related hashtags organize and anchor the content they describe and, reciprocally, any regularities in the content associated with a hashtag cement the meaning of its referent.
Investigating the construction of values on Instagram: a conceptual roadmap As mentioned in the "Introduction" section, the concept of values has not been foregrounded in studies of Instagram. This is rather striking given the widespread moral panic over the alleged lack of values among Instagram users (Miltner & Baym, 2015), along with contested attempts to link Instagram with personality traits in general and narcissism in particular (e.g., Sheldon & Bryant, 2016). Furthermore, a lot of Instagram research deals with value-laden topics such as gendered selfpresentation (Baker & Walsh, 2018) and humanitarianism (Sin & He, 2019).
In the absence of a corpus of literature that specifically focuses on values on Instagram, our analysis finds theoretical grounding in a combination of established value theories and studies about Instagram as a platform. A joint reading of this work led us to identify two fundamental tensions that may facilitate the interpretation of the visualization of values on the platform. The first one is the relatively universal tension that underlies the relationship between individuals and social groups. Based on value theory (e.g. Schwartz, 1992), we summarize this as the duality between individualism and collectivism. This tension provides us with a broad framework to investigate the representation of values in images taken at different points in time and space. The second tension pertains to contemporary modes of civic participation. Given the role of social media platforms in present-day societies (e.g., van Dijck, Poell, & de Waal, 2018), we summarize this tension in terms of the duality between liberal and neoliberal modes of participation. Focusing on this tension allows us to situate our study within key debates over the role of social media in the present socio-political moment.
The individualism-collectivism duality has been widely explored in veteran value theories, from political science (Inglehart, 1971) to organization studies (Hofstede, 1980). Individualistic and collectivistic values have been traced back to the tension between two basic needs that are essential for human survival. On the one hand, humans heavily depend on group welfare and cooperation, yet on the other hand, they wish to satisfy their personal biological needs as well as their desire to understand, explore, and transform their environment (Schwartz, 1992).
For the purpose of analysis, and with awareness to the over-simplification often associated with these axes, we interpret the individualism-collectivism duality as value orientations toward the "self" or "others." The concepts of "networked individualism" (Wellman et al., 2003) and "networked self" (Papacharissi, 2010) speak to the relevance of this duality to social media in general and Instagram in particular. People go to Instagram to communicate their own perspective on life through a complex performative process in which they present different parts of their subjectivity through images (Zappavigna, 2016). The individual constructed via social media, however, is also inevitably "networked," as self-presentation facilitated by the platform pivots around social connections. Thus, people use Instagram to conduct markedly social activities such as collective self-disclosure (Das & Hodkinson, 2019) or ritualized mourning (Burgess, Mitchell, & Munch, 2018). Instagram is thus a place where people construct their own self together with their respective others.
The second duality through which we examine values on Instagram is the logics of liberalism and neoliberalism as embodied by the figures of the "citizen" and the "consumer." Here, we draw on Wendy Brown's (2015) ideal types of homo politicus (the political person) and homo oeconomicus (the economic person). Homo politicus embodies the ideal citizen in a liberal democracy: a sovereign subject who enacts their freedom by engaging with their peers in the public forum in pursuit of the common good. In contrast, the homo oeconomicus finds value exclusively in financial accumulation, striving to maximize their returns on investments in all areas of life. This type of market orientation may be expressed through selfcommodification and consumption. For example, the reaction of the citizen to a societal problem such as climate change might take the form of collective action demanding restrictions on carbon emissions, while the reaction of the consumer might involve a shift in consumption patterns toward eco-friendly products.
The citizen-consumer duality is arguably central to the study of social media platforms, which operate in a system where capitalism and democracy represent clashing sets of values (van Dijck et al., 2018). Despite the documented use of Instagram for political activism (among many others, see Einwohner & Rochford, 2019), the platform tends to promote so-called "commodity activism" (Mukherjee & Banet-Weiser, 2012) in the wider context of a life that must be "shoppable" (Hund & McGuigan, 2019). For example, while there is an active body positivity movement on Instagram (Cohen, Irwin, Newton-John, & Slater, 2019), brands and corporations exploit its visibility to engage in direct advertising, effectively silencing activists and reshaping the structure and goals of the movement itself (Cwynar-Horta, 2016).
Drawing on these bodies of knowledge about Instagram, values, and visualization, we seek to explore the overarching question of which visual repertoires are associated with value-related terms on Instagram. We do this through an analysis of two interrelated aspects. First, we trace the breadth of these repertoires, namely their openness to multiple interpretations or their specific character. Second, we investigate how these images construct the meaning of values in light of the two sets of dualities outlined above.

Methods
To address the complex issues presented thus far, we inductively and collaboratively analyzed a large corpus of Instagram images using mixed methods. We outline the eight steps of the process in Figure 1.
Step 1: Screening of value-related search terms. An international team of six researchers from six countries (Germany, Israel, Italy, Japan, South Korea, and the United States) screened potential search terms based on Christen, Narvaez, Tanner, and Ott's (2016) list of value-related terminology, consisting of over 400 words with demonstrated cross-cultural salience. Out of pragmatic necessity, we screened search terms in English only. Nonetheless, since we expected the corpus to be transnational in light of the literature on the globalization of social media platforms (Jin, 2019), we strove to select value words that would resonate across cultures. Applying the principles of consensual qualitative research (Hill, Thompson, & Williams, 1997), the team came to a strong agreement on the salience of 57 terms across their different language communities. Examples of the values in the list include beauty, compassion, equality, health, and freedom.
Step 2: Selection of Instagram search terms. We assessed each term's popularity as an Instagram hashtag based on the number of associated posts. At this stage, we What Does #Freedom Look Like?

882
Journal of Communication 71 (2021) 875-897 discarded several hashtags that anchored images referring to people's attitudes toward objects rather than to values (e.g., "I love New York City"). Our final list included the top 20 value-related terms (outlined in Table 1), ranging from 36þ million posts for the most popular hashtag in the sample (Beauty) to 5þ million posts for the least popular one (Sustainability).
Step 4: Ordering, cleaning, and capturing the corpus. We sorted the subsets according to the number of likes and filtered out multiple posts from the same account. This data cleaning strategy minimized systematic hashtag spamming by businesses and influencers. Thereafter, a team of four coders cleaned the data, removing all posts that did not feature the relevant hashtag in the caption or comment by the original author. The coders then took screenshots and archived the top 100 pictures based on the number of likes for each of the 20 value terms. The final dataset thus contained 100 entries for each of the 20 subsets, leading to a total of 2,000 images and their metadata. An initial inspection of the images revealed the transnational nature of the corpus, encompassing various nationalities, locations, and languages.
Step 5: Quantitative coding. The 2,000 images were coded by a team of four members over a one-month period. The codebook was inspired by Kress and van Leeuwen's (1996) toolbox for visual analysis, which offers usable descriptors of the common elements of visual compositions (the authors will happily share the codebook upon request). After a pilot phase, the coders performed a reliability test on a set of 50 units randomly sampled from the 20 subsets, passing Krippendorff's alpha threshold for intercoder reliability in exploratory studies (a ¼ .67) for all variables.  Specifically, coders were asked to identify the protagonist of the representation (a ¼ .90); the main setting (a ¼ .69); the distance between represented participants and the viewer (a ¼ .83); and whether the picture was a selfie or not (a ¼ .89). For example, in the "protagonist" variable, coders were asked to determine whether the main subject of the representation was an individual, group, animal, object, built environment, natural environment, text, artistic representation or something else. Following the intercoder reliability test, the 2,000 units comprising the final dataset were coded.
Step 6: Statistical analysis. The association between each value and the four variables was tested using Fisher's exact test, corrected for multiple testing using the Benjamini-Hochberg false discovery rate procedure. This analysis allowed us to identify unique features of the 20 subsets in comparison to the rest of the data.
Step 7: Formulation of empirically sound ideal types. Once we identified the salient features of the 20 value repertoires, we moved on to the qualitative phase of the analysis. First, we developed short descriptions of the ideal types for 19 of the 20 value words, combining insights from the quantitative and qualitative analyses. As described below, one value (Gratitude) was not associated with a clear visual repertoire, and we decided to discard it in the advanced analytical phases. 1 Alongside this process, we examined the metadata of the sample to locate the most frequent cooccurring hashtags for each of our search terms.
Step 8: Descriptive and interpretive axial coding. Finally, and inspired by the distinction between the iconographic description of visual motifs and the iconological interpretation of images in relation to their broader socio-historical context, we conducted an integrative analysis of the entire sample at two levels: descriptive and interpretive. At both levels, we used the principles of grounded theory (and with it, axial coding) to chart the relationships between abstract categories and our data (see Lindlof & Taylor, 2019). The process involved a combination of individual coding, constant comparison, and ongoing deliberation and discussion among members of the multi-lingual and multi-national research team.
For what concerns description, the team surveyed the corpus several times, seeking to detect patterns in the sampled images. This led to the identification of two dualities around which we structured our analysis: showing-telling and nature-culture. The former refers to the sign-type dominating the representation (icons or text). The latter refers to the dominant feature of the setting (nature or the built environment). We see these as short-hands for empirically derived descriptors that can summarize and convey the content of the images in a manner that is meaningful to the analysis as well as intelligible to the reader. Thereafter, we located the 19 values on an imagined plane created by the intersection of the two axes.
For what concerns interpretation, we sought again to stay as close to the data as possible while also offering theoretically relevant insights based on the aforementioned axes of self-other and citizen-consumer. In line with later developments in the grounded theory approach (see Kelle, 2007), our analysis combined previous theoretical groundings with continuous inspection of the data. As was the case for the descriptive analysis, we concluded this phase by positioning each ideal type on the imagined plane created by the intersection of the two axes.

Value repertoires: an overview
The most striking finding of the quantitative analysis was that, except for one, each of the values in our sample possess a distinct visual identity. In other words, the value repertoires are internally consistent in terms of the participants represented, the setting, and/or the relation to the viewer. For each value, the statistical analysis revealed at least one distinguishing attribute. Table 1 details the objects and events specific to each repertoire.
Further qualitative analysis allowed us to define and detail the visual motifs constituting the ideal types for each value repertoire. This was corroborated by our analysis of co-occurring hashtags. In all cases, the hashtags that frequently cooccurred with each of our value terms cohered with the motifs that we identified through our visual analysis. It is important to note that the ideal types do not account for the totality of the sample and outliers were present to varying degrees. However, images resembling the statistically derived ideal types were highly prominent across the sample (see Figure 2). After this initial typification, we further analyzed the sample to detect patterns of similarity and difference between the various image repertoires. This analytical phase led us to cluster the visual repertoires into five main groups.
The first cluster includes the value repertoires of Freedom and Happiness, which feature individual human protagonists outdoors and away from built environments. We name this cluster "alone in nature." Freedom is typically represented by medium-long shots of individuals looking into the distance, often outdoors. Occasionally, representations gathered under this hashtag also featured panoramic shots of a natural landscape, often heavily filtered. Happiness typically featured close shots of individuals looking into the distance, surrounded by nature.
The second cluster is the largest in our dataset and gathers together value repertoires that feature individual human protagonists indoors, often exercising. Tagged with #flexibility, #health, #strength, and #progress, we name this cluster "gym bodies." Flexibility is represented through medium-long shots of individuals practicing some form of stretching. Health and Strength are represented through mediumlong shots of individual people, often dressed to reveal toned muscles. Finally, Progress features mirrored selfies taken from a distance to show the full body of an individual, again dressed to show off toned muscles.
The third cluster of value repertoires features text as the preferred mode of expression. Tagged with #faith, #mindfulness, #spirituality, and #knowledge, we name this cluster "cultivating the mind." The values in this cluster are dominated by short pieces of writing imposed over plain backgrounds, dark backgrounds featuring a range of symbols, or text-image combinations where the text is dominant. Faith favors text about religious inspiration and identifiable religious symbols. Mindfulness and Spirituality do not share this overtly religious orientation, favoring text with a wider spiritual appeal. Knowledge features image-text combinations of bite-sized pieces of trivia alongside stock photos. As opposed to the other values in this cluster, the aesthetics of Knowledge are not pristine and reflexive, but quite colorful and almost deliberately "ugly" (Douglas, 2014).
The fourth cluster is composed of the value repertoires of Loyalty and Respect, which present groups of people as their main subject. We name this cluster as "performing relationships." Both values feature photographs of families and friends, usually in the form of a medium shot emphasizing relationality among the participants represented.
The fifth and final cluster features value repertoires that focus on human-made objects. These are the ones tagged with #competition, #creativity, and #sustainability, a cluster that we name "objects on display." Competition gathers representations Figure 2 Artist's impression of Instagram images closely fitting each of the ideal types (the order of presentation follows the description in the paper). Art provided by Blake Hallinan. of manufactured objects set in an urban environment as their main subject, often for advertising purposes (e.g., giveaway competitions). Sustainability also features representations of manufactured objects presented in an earth-tones aesthetic for advertising purposes, suggesting that the object was produced "sustainably." To a lesser extent, this value repertoire also features representations of sustainable living facilities (e.g., tiny houses) and holiday destinations (e.g., eco-friendly hotels). Creativity features close-up representations of art pieces, both analog and digital.
The remaining value repertoires have distinct watermarks but do not cluster as neatly as the ones described above. Beauty gathers close-up selfies of individuals gazing directly at the camera. A smaller proportion of pictures in this value repertoire also feature the built environment (e.g., a city's skyline, monuments), often heavily filtered. Tradition features highly-edited images of the built environment and, to a lesser extent, pictures of individuals wearing "traditional" outfits or consuming "traditional" food. Compassion is almost equally split between representations of animals and text-based representations resembling those of #mindfulness or #spirituality. Equality features text or image-text combinations referencing mainly LGBTQþ rights (e.g., statements about marriage equality accompanied by a rainbow flag).
A conceptual map of value repertoires As described above, a grounded analysis of the recurring visual motifs in the corpus led to the identification of two conceptual axes: nature-culture and showing-telling. While the "nature" end of the spectrum features pictures set in nature or in which nature is the main subject, the "culture" end features pictures located in built environments or where the main focus of the representation is either a built environment or a human-made object. The second axis seeks to capture the duality between icons and symbols as the preferred sign type. The end we name "showing" features pictures (icons) and the end dubbed as "telling" focuses on text (symbols). The quadrant chart below (Figure 3) organizes the values in our corpus around these dualities.
As evident from the graph, the majority of values are located in the "showingculture" region. These feature icons as their preferred sign types and depict cultural artifacts or relations between human subjects. Competition, Sustainability, and Creativity have manufactured products or art pieces as the protagonist. Tradition focuses on the built environment, traditional outfits, or traditional food. Loyalty and Respect foreground communality among human participants as members of the same cultural unit (e.g., family and ethnicity). Flexibility, Health, Strength, and Progress feature bodies shaped by gym culture, while Beauty presents bodies modified by make-up. We interpret these to be "cultured bodies," that is, bodies modified by cultural practices.
The repertoires of Happiness and Freedom show human protagonists set in nature. They emphasize the individual experience of being outdoors, even when the built environment happens to be in the frame in one form or another (e.g., a beachside shack). As discussed below, the meanings implied by this setting suggest a desire to distance one's self from civilization.
Finally, the value repertoires of Mindfulness, Faith, Spirituality, and Knowledge show a strong preference for articulation through text. On a platform strongly associated with photography, the presence of these text-based repertoires is striking. We see this as a cue which suggests an inherent difference in how people understand and represent values. While the "gym bodies" cluster tends to emphasize the final results of an embodied process, the text-based values in the "cultivating the mind" cluster appeal to a continuous process of internal work. We further discuss the theoretical implications of this distinction in the concluding section.
Moving to interpretation, we conducted a theoretically informed analysis of the data based on the aforementioned axes of self-other and citizen-consumer (see Figure 4).
The "self-citizen" quadrant of the graph is remarkably empty. None of the value repertoires featured an orientation that sees the individual as an agentic member of a political community. While some of the Compassion images do address individuals as citizens, this value repertoire is split between the other-orientation of animal rights activism and the self-orientation of inspirational posts encouraging people to be compassionate toward themselves.
Equality is the only value repertoire that we placed in the "citizen-other" quadrant, as it gathers politicized messages in support of the enfranchisement of the LGBTQþ community. While such an orientation might be expected for a set of images anchored by the term "equality," the repertoire is still surprisingly narrow, as other common (in)equalities such as race and class are notably absent. While discussion of race and class is certainly present on Instagram, the fact that the broad term "equality" is narrowed down to a singular issue is striking. The value repertoires of Loyalty and Respect fall on the "other" end of the graph because of their orientation toward group belonging. However, they fall in the middle of the vertical axis because they suggest no commitment to either active citizenship or consumption. The "other-consumer" quadrant features the value repertoires of Tradition and Sustainability. Tradition reflects a consumer-orientation, as it focuses on "traditional" food and manufactured goods for purchase, but is also otherdirected because of its focus on group identity. Sustainability foregrounds its commercial character despite being framed as environmentally-conscious and therefore altruistic. Different forms of corporate greenwashing tagged with #sustainability include the promotion of products, holiday destinations, and lifestyle products in earth tones.
The majority of the value repertoires fall in the "self-consumer" quadrant. With its strict focus on Instagram giveaways, Competition has a clear consumer orientation. Somewhat surprisingly, the same applies to Knowledge. Indeed, this value repertoire features bite-sized pieces of trivia that individuals can easily consume and digest, backgrounding the wider process of knowledge acquisition and the possibility of using knowledge for the public good. The value repertoires of Flexibility, Health, Progress, and Strength are heavily self-oriented in light of their focus on the individual body. They also have an orientation toward the consumption of goods and services necessary to cultivate the body (e.g., healthy food and gym subscriptions). The same applies to the value repertoire of Beauty, with its emphasis on the self and orientation toward the consumption of make-up products.
We interpret the value repertoires of Happiness and Freedom as heavily selforiented due to their focus on an individual's ability to escape modern life and find refuge in nature. The protagonists celebrate the fact that they are alone by literally "freeing" themselves of clothes, dancing in the middle of the street, or just enjoying a moment of solitary contemplation. Positioning these representations on the citizen-consumer axis proved to be difficult, as two competing interpretations emerged from the data. For both values, some representations are clearly oriented toward the consumption of time alone in nature. For example, outdoor vacation pictures are common in both repertoires. Just as commonly, however, our images do not include any indication of a consumerist ethos. For example, shots of individuals on a solitary mountain hike might suggest a wish to move away from civilization, including consumerism. Often enough, it was impossible to tell these two orientations apart. Accepting both interpretations as valid, we placed Happiness and Freedom in the middle of the citizen-consumer axis.
Finally, we interpret Creativity, Faith, Mindfulness, and Spirituality to be heavily self-oriented value repertoires because of their focus on individual self-expression, inner balance, and self-love, respectively. Even when the messages directly address other people, the prescribed actions ultimately relate to the self rather than to others. For all four value repertoires, however, we found no clear orientation toward either active citizenship or consumption.

Concluding remarks
So, what does freedom look like? At least on Instagram, #freedom typically features an individual on the beach or a mountaintop, enjoying a moment of solitary contemplation. Freedom and the other values are abstract concepts, theoretically open to multiple forms of visualization, especially when these visualizations are crowdsourced by individual social media users. Yet, we hypothesized that such value imaginaries would be bounded by the logics of the hosting platform, and explored the nature and properties of those boundaries on Instagram. An analysis of the images associated with 20 popular value-related hashtags revealed that almost every value had a distinct visual footprint. Overall, we demonstrated that Instagram works as an archive that organizes its content around the folksonomy of hashtags, in which the images contributed by users form paradigms that constrain the visualization of their referents.
We also found that the value repertoires in our sample expressed a strong orientation to the self. This finding is congruent with previous studies that depict social media as spaces where people engage in the networked construction of personal identity. Furthermore, the sample is characterized by a general orientation toward consumption, with most values visualized as either apolitical or overtly commercial. Overall, our analysis supports a reading of Instagram as a "storefront" (Hund & McGuigan, 2019) where people go to access products (e.g., Competition) or perform their consumer identities (e.g., Beauty). We might have found different framings if we had investigated narrower or more overtly political terms, yet the finding that broad concepts such as "freedom" or "progress" are de-politicized in this sphere is still illuminating.
Based on our results, we offer three interrelated conclusions which pertain to image-text relations in the production of the meaning of abstract concepts, the distinction between the representation of internalized and externalized values, and the role of aestheticized consumption as a core principle shaping Instagram content. Our conclusions, we wish to restate, are based on a corpus composed of public, highly visible images. The generalizability of our observations, therefore, applies to the mainstream culture of Instagram.
Our first conclusion relates to the complex relationships between concepts and images. In our analysis, we focused on the transition from the former to the latter, claiming that user-generated content is archived on Instagram as bundles of related images pegged to hashtags. In the process, the potentially endless visualization of a term gets narrowed down to a specific, iconic template. However, the transition is anything but unidirectional. Images can and do act back on their referents, influencing their meaning and defining their scope. Originally anchored by their associated search terms, recurring visual motifs reshape the terms themselves, narrowing down, diluting, or displacing their meaning. Hence, Health on Instagram is narrowed down to muscle gains, Knowledge is diluted to only signify trivia, and the political salience of Freedom is displaced in favor of solipsistic contemplation. Thus, images posted by users contribute to temporarily "fix" the meaning of value-related terms by establishing iconic templates for them.
Secondly and connectedly, we observe that the process through which images and their referents influence each other can take very different forms. The main distinction we offer in this respect is between values that are externalized through representation via objects (including the human body) and values that are internalized through textual representation. Generally speaking, externalized representations adopt aesthetic canons with broad social validation and prestige (e.g., conventionally attractive bodies), while internalized representations choose text to convey a generic and disembodied call for self-transcendence. However, it is important to stress that these visual patterns are not inherently tied to the value terms in our sample. A value externalized through representations of muscular bodies such as Strength could in theory be internalized through text-based messages inviting individuals to cultivate their resilience in the face of adversity. Conversely, an internalized value expressed via text such as Mindfulness could in theory be externalized through pictures of yoga instructors exercising with an ocean view in the background. It is therefore clear that this distinction emerges from the hashtag-specific vernaculars of our sample and the representational practices of Instagram users. Furthermore, the distinction between externalized values associated with social prestige and power and internalized ones that call for self-restraint and transcendence may also be relevant to how values are visualized on social media platforms other than Instagram.
Finally, within the realm of the embodied, externalized values, we argue that aestheticized consumption is a guiding principle shaping the aesthetics of Instagram's mainstream. Simply put, this is a value orientation towards the aestheticization of everyday life through consumerism. The concept resonates with the original meaning of the term "lifestyle," invented to identify consumer-types based on their values, attitudes, and behavioral patterns for market segmentation (Mitchell, 1983). Through the lens of aestheticized consumption, core values such as the ones in our sample are narrowed down and transformed into visual representations of goods for consumption. Sustainability becomes visualizable mainly in terms of green marketing; Tradition becomes associated with products like costumes and food; and Health doesn't involve medical doctors or surgical masks, but gym subscriptions and green smoothies to achieve a toned body.
This analysis suggests that the narrower meaning of value-related terms is not random, but follows a consumption-based paradigm in which things have to be marketable and monetizable in order to achieve enough visibility to enter the mainstream of Instagram's archive. A different way to look at this is to say that the commercial logic of Instagram's mainstream turns values into positively charged signifiers and puts them to work to increase the market appeal of specific goods and services. A poignant example is that of overtly commercial posts (e.g., corporate advertisements) bundling together multiple value-related hashtags such as #happiness, #freedom, #success, and #progress to boost engagement.
Interestingly, however, this tendency to use value-hashtag bundles involves corporate-, influencer-, and user-generated content alike. The implication might be that aestheticized consumption is more than a marketing strategy, but actually one of the core values around which Instagram's archive is organized, so much so that it influences the aesthetic choices of private users, re-orienting the visualization of other values on the platform. This amounts to saying that the visualization of values in Instagram's archive is tied to pre-existing structures such as the top-down logic of market value, even though the images composing it are bottomup contributions.
Before concluding, we wish to point to some of the limitations of this study that will hopefully be addressed in future work. First, our list of values does not cover the whole range of values that could and should be explored. Moreover, these hashtags are deliberately broad and generic. It is entirely possible that the narrowing down process described above does not apply in the case of more specific political hashtags such as #BLM. Narrower search terms might lend themselves to an "opening up" rather than a narrowing down of meaning. Such hashtags may also foreground a political subjectivity that was surprisingly absent from our corpus.
Secondly, while Instagram may well be the place for the visualization of everything, it is by no means the only place where things are visualized. Instagram is not representative of the full social media landscape, let alone visuality at large. Furthermore, we focused on a specific facet of Instagram: still images in publicly available and highly visible posts. Thus, our conclusions may well not be relevant to the interpretation of the visual repertoires circulated in more peripheral or private parts of Instagram's archive. These might follow different logics and merit further research. Thirdly, we acknowledge the Anglo-centrism of our search terms, even though the images themselves appeared to originate from many locations worldwide. We are in the process of complementing this perspective with several cross-linguistic follow-up studies.
Finally, while our results are inevitably bound to the time frame of data gathering, we were surprised to discover that images associated with general value hashtags seem to be less affected by specific events than one might expect. Consider the impact of COVID-19 on the images in our sample. While the pandemic rocked the world during our data collection period, explicit impact on our data was minimal. Most surprisingly, the posts associated with Health hardly mention the virus, preferring the abovedescribed focus on gym-toned bodies. This is not to suggest that we see the value repertoires emerging from our data as stable and unchangeable-a claim that would be incompatible with our epistemology. Recognizing the socially constructed and therefore contingent character of our results, we invite future studies that replicate our protocol to analyze potential shifts in the visualization of values over long periods of time.
Limitations notwithstanding, our work sheds light on key questions related to the visualization of values on Instagram. Empirically, we defined the visual repertoires associated with prominent values, mapped how they relate to each other, and analyzed the role of aestheticized consumption as a core principle reorienting how users visualize values. Methodologically, we offered a comprehensive strategy for analyzing the visual representation of value hashtags on social media. Conceptually, we showed how visual repertoires used to represent core values construct specific (and often narrow) meanings for their referents. In addition, the distinction between internalized and externalized constructions of values, in which we showed how value repertoires emphasizing self-transcendence are heavily text-based and those emphasizing self-enhancement are predominantly image-based, provides a conceptual roadmap for understanding the multi-modal communication of morality. Altogether, this study demonstrates that the analysis of images offers a primary avenue for exploring the expression of values on social media, significantly expanding our understanding of the social construction of values in digital spheres.

Funding
This project has received funding from the European Research Council (ERC) under the European Union's Horizon 2020 research and innovation programme [Grant Agreement No. 819004].